Using artificial neural network to predict the frequency of three-dimensional frame structure

Abstract

The paper applies the ANN model with Leveneberg - Marquardt, Bayesian Regularization, Scaled Conjugate Gradient techniques to predict the frequencies of the three-dimensional frame structure. The computational model of the problem is a frame - pile structural system in the form of three-dimensional frame, linear elastic deformation, pile - soil link is replaced by a hard restraint (fixed) with equivalent restraint depth. With almost accurate prediction results (with an error of no more than 1.58%), it shows that the application of ANN with the above techniques in predicting the frequencies of the three-dimensional frame structure is reliable, feasible. This is very meaningful in solving problems of identification and diagnosis of structures, especially with complex structures, large of problems, parameters affected by many changing factors in the process of exploitation, suitable to areas with difficult conditions in weather, climate and difficult conditions in information technology infrastructure and specialized software.

Key words: Frequencies; predict; Leveneberg - Marquardt; Bayesian Regularization; Scaled Conjugate Gradient.